Different patterns can be recognized in EEG recordings preceding
the onset of movements. At least two different processes can be
distinguished: event-related de-synchronization (ERD) and
synchronization (ERS), and movement related potentials (MP).
The thesis project will focus on the comparison between
the prediction of movements based on ERD/ERS and MP. It will
address the problem of detecting classification differences in
performance between the two kind of processes, finding a good
timing for the detection of movements, and trying to understand
whether it might be convenient to combine both patterns in order
to predict the movement.
Most part of the work will focus on what concerns ERD/ERS patterns.
The workflow will contain: literature search for established methods
to detect movements based on ERD/ERS; implementation in DFKI software
framework (programming language Python); evaluation and comparison
with MP based prediction and if applicable evaluation of the
combination of the two approaches. The work will focus mainly on
pre-processing and feature generation, but not so much on
classifiers development. The latter could be taken into account
if it results required for combining the two kind of patterns.